Today, our autonomy compute hardware is general-purpose, power-intensive, and heavy. The Platform Subteam will address these shortcomings, and investigate the burgeoning possibilities of heterogeneous computing, leveraging industry FPGA SoCs.
We hope to develop and integrate optimized computing platforms to improve autonomy capabilities and achieve research outcomes under the guidance of faculty. We plan to use a mix of traditional CPU computing with FPGAs to take advantage of their core advantages:
- Parallelization: FPGA’s parallel computing capabilities allow us to accelerate large-scale operations, such as image processing, and assign independent tasks to dedicated hardware sections.
- Configurability: FPGA’s configurability allows us to iteratively improve our self-driving algorithms and easily synthesize new designs. Their customizable I/O blocks allow us to optimally interface compute with exteroceptive sensing for self-driving.
If this challenge is of interest to you, please reach out to Full Team Lead Kunal Gupta (kg379) directly.